• Can we call this eminence based medicine?

    Normally I’d post a reading list at this time, but my usual help in putting one together is unavailable this week and next. So, you’re on your own. Sorry. In its place, here’s a post based on a paper Aaron drew my attention to. 

    In a 2011 paper, members of the US Preventative Services Task Force (ungated pdf) explained how the organization evaluates services in general and how it came to its decision about lipid screening for children in particular. The clear message delivered by the authors is that the USPSTF process is nothing if not systematic and transparent, incorporating peer-review and public input, and including a rigorous check for conflicts of interest. In the case of lipid screening for kids, the body concluded that the evidence was insufficient to recommend for or against screening. Consequently, according to the USPSTF, any screening advice, for or against, is not evidence based.

    Because the evidence is inconclusive, the authors suggest that a shared decision making (SDM) process be employed to help clinicians decide with their patients whether to screen, consistent with patients’ (or, in this case, probably their patients’ parents’) values. The authors also note that in the absence of an official recommendation, there is great opportunity for research. In fact, before practice patterns are established is the best time for randomized trials. After practice is standardized — whether based on evidence or not — randomized trials might be considered unethical. Observational studies are more difficult when practice patterns are less varied and possibly correlated with unobservable quality. (On this point, related posts are here and here.)

    However, in the case of lipid screening for children (and perhaps other tests for which the USPSTF finds inconclusive evidence), there is another organization that is willing to make a recommendation. The American Academy of Pediatrics (AAP) does recommend lipid screening for certain kids. The authors take issue with the AAP’s process by which that determination was made.

    The original recommendations by [AAP's] National Cholesterol Education Program were based on those of an expert consensus panel, in which scientific and clinician experts participated in a commonly used process to arrive at clinical guidelines. As with many expert panels, the methods used to identify and analyze scientific evidence were not explicitly defined in the report. In its subsequent updated clinical reports, the Committee on Nutrition did not reference or outline a clear method for its updated literature review. [...]

    Because this AAP clinical recommendation appears to have been based on less systematic and transparent methods, users may not know whether the recommendation is consistent with the evidence, whether guideline developers factored in important research gaps when developing the recommendation, or how potential conflicts of interest may have been managed.

    To the extent the AAP process is not evidence based (and it is to some extent, but it is hard to tell how), what is it? Let’s call it eminence based, or at least partly so. Does that seem fair? If not, why not? Can you think of a better example of eminence based medicine?


    Comments closed
    • Do you want a better example, or just one of the (many) examples — and species — of eminence-based medicine?

      Here’s what I mean by species: Some of what you’re calling eminence-based medicine is based on virtually no science whatsoever, or very little valid (i.e. likely to be true), or high-quality evidence. The guidelines for pediatric cholesterol screening are one example. Other examples include orthognathic surgery for TMJ, spinal fusion for uncomplicated back pain due to ruptured disc; the use of vena cava filters to prevent pulmonary embolism; proton beam therapy for prostate cancer. I would argue that the use of flu vaccine also falls into this category, but that’s controversial.

      Some medical services are done even in the face of <> evidence. In other words, the evidence says don’t do it because it doesn’t work. Vertebroplasty, for instance. Pulmonary artery catheters.

      And some treatments are based on solid evidence, but they are given to patients who are either highly unlikely to benefit, or who are entirely inappropriate patients, e.g. giving a stent to a patient who has no symptoms of angina, or giving Herceptin to a breast cancer patient whose cancer is not estrogen-positive (which means it will not be susceptible to the drug).

      All 3 species described above are “credence goods.” They are purchased on the basis of trust in an expert.

      • I gotta say, I think there is a big difference between the vena cava filter evidence and the children cholesterol screening evidence. Vena cava filters have been studied in patients who have had a PE, and found at 8 years to provide a very mild (6% I think…) reduction in rate of recurrent PE with no benefit to overall mortality.
        Some may say that the study was too small (possibly…), or the time period too short (I would disagree here, it was an elderly population, and 8 years is a pretty good time to measure a mortality benefit in an older crowd…), but there is evidence that basically says that vena cava filters kind of maybe work, but really dont’ do much….

        That is in contrast to cholesterol screening for kids, where the major question that needs to be answered is “in 30-40 years, do kids who were screened for high cholesterol have less CV events or even less mortality than those who weren’t.” In this instance, it’s more that the study really really has never been done…. (and probably never will be…) so we rely on extrapolation from what we do know:
        1) there is a rising percentage of children with unfavorable lipid profiles
        2) those children have a high rate of abnormal lipid profiles as adults
        3) we believe that long term elevated cholesterol increases your risk of heart attack, stroke, peripheral vascular disease, etc. Though the effects may not be realized for years down the road.

        So yeah, this is not evidence based medicine, but I think there has to be a question of if there can ever really be a high quality (double blind RCT) that answers the question being asked…

    • I would hope that expert opinion (eminence-based opinion) is closely correlated with the evidence. Where that is not the case I wonder if the experts are overvaluing personal experience–perhaps experience that hasn’t translated into published research– or the rationale, as opposed to the evidence, for a procedure. I think that one key difference between researchers and other experts is that researchers have a bias in favor of the null hypothesis while most people and most practitioners have a bias against the null hypothesis. We want to believe that things to work. I think that bias may partially explain some of the tendency to overrate the importance of survival rates noted earlier. if an increased survival rate with a new test COULD be due to using test, that is good enough for many practitioners